Databricks Machine Learning Associate Question 575
Select 2You are training a machine learning model on Databricks to predict customer churn. You want to optimize the model's performance by tuning its hyperparameters. You decide to use hyperparameter search. Which of the following statements about random search, grid search, and Bayesian search are correct?
- A
Random search can often find good hyperparameter combinations more efficiently than grid search.
- B
Grid search is more efficient than random search when the hyperparameter space is large.
- C
Bayesian search uses prior results to guide the selection of the next hyperparameter set to evaluate.
- D
Random search is guaranteed to find the best hyperparameter set if given unlimited time.
- E
Bayesian search is computationally cheaper than grid search in all cases.
Show answer and explanation
Correct answers: A, C
Explanation
Random search is often more efficient than grid search in exploring large hyperparameter spaces because it does not require exhaustive evaluation of all combinations. Bayesian search is an advanced method that uses prior evaluation results to guide further exploration, making it more sample-efficient than random or grid search. However, neither random search nor Bayesian search is universally optimal, and the choice of method depends on the model complexity and available computational resources.
- A. Correct.
Random search is often more efficient than grid search because it explores the hyperparameter space more diversely and can find good combinations without testing every possibility.
- B. Incorrect.
Grid search becomes computationally expensive and inefficient when the hyperparameter space is large, as it exhaustively evaluates all possible combinations.
- C. Correct.
Bayesian search builds a probabilistic model based on prior evaluations and uses that to guide the selection of the next hyperparameter set, making it more efficient.
- D. Incorrect.
Random search improves coverage of the hyperparameter space but is not guaranteed to find the best hyperparameter set, even with unlimited time.
- E. Incorrect.
Bayesian search can be computationally expensive, especially for models with high training time, and is not universally cheaper than grid search.